2 * Copyright (c) 2003,2004 The DragonFly Project. All rights reserved.
4 * This code is derived from software contributed to The DragonFly Project
5 * by Matthew Dillon <dillon@backplane.com>
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
11 * 1. Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 * notice, this list of conditions and the following disclaimer in
15 * the documentation and/or other materials provided with the
17 * 3. Neither the name of The DragonFly Project nor the names of its
18 * contributors may be used to endorse or promote products derived
19 * from this software without specific, prior written permission.
21 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
22 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
23 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
24 * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
25 * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
26 * INCIDENTAL, SPECIAL, EXEMPLARY OR CONSEQUENTIAL DAMAGES (INCLUDING,
27 * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
28 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
29 * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
30 * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
31 * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
34 * Copyright (c) 1990 The Regents of the University of California.
35 * All rights reserved.
37 * This code is derived from software contributed to Berkeley by
40 * Redistribution and use in source and binary forms, with or without
41 * modification, are permitted provided that the following conditions
43 * 1. Redistributions of source code must retain the above copyright
44 * notice, this list of conditions and the following disclaimer.
45 * 2. Redistributions in binary form must reproduce the above copyright
46 * notice, this list of conditions and the following disclaimer in the
47 * documentation and/or other materials provided with the distribution.
48 * 3. All advertising materials mentioning features or use of this software
49 * must display the following acknowledgement:
50 * This product includes software developed by the University of
51 * California, Berkeley and its contributors.
52 * 4. Neither the name of the University nor the names of its contributors
53 * may be used to endorse or promote products derived from this software
54 * without specific prior written permission.
56 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
57 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
58 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
59 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
60 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
61 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
62 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
63 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
64 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
65 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
68 * $FreeBSD: src/sys/i386/i386/swtch.s,v 1.89.2.10 2003/01/23 03:36:24 ps Exp $
69 * $DragonFly: src/sys/platform/vkernel/i386/swtch.s,v 1.8 2007/07/01 02:51:43 dillon Exp $
74 #include <sys/rtprio.h>
76 #include <machine/asmacros.h>
77 #include <machine/segments.h>
79 #include <machine/pmap.h>
80 #include <machine/lock.h>
85 #define MPLOCKED lock ;
94 #if defined(SWTCH_OPTIM_STATS)
95 .globl swtch_optim_stats, tlb_flush_count
96 swtch_optim_stats
: .long 0 /* number of _swtch_optims */
97 tlb_flush_count
: .long 0
104 * cpu_heavy_switch(next_thread)
106 * Switch from the current thread to a new thread. This entry
107 * is normally called via the thread->td_switch function, and will
108 * only be called when the current thread is a heavy weight process.
110 * Some instructions have been reordered to reduce pipeline stalls.
112 * YYY disable interrupts once giant is removed.
114 ENTRY
(cpu_heavy_switch
)
118 movl PCPU
(curthread
),%ecx
119 movl
(%esp
),%eax
/* (reorder optimization) */
120 movl TD_PCB
(%ecx
),%edx
/* EDX = PCB */
121 movl
%eax
,PCB_EIP
(%edx
) /* return PC may be modified */
122 movl
%ebx
,PCB_EBX
(%edx
)
123 movl
%esp
,PCB_ESP
(%edx
)
124 movl
%ebp
,PCB_EBP
(%edx
)
125 movl
%esi
,PCB_ESI
(%edx
)
126 movl
%edi
,PCB_EDI
(%edx
)
128 movl
%ecx
,%ebx
/* EBX = curthread */
129 movl TD_LWP
(%ecx
),%ecx
130 movl PCPU
(cpuid
), %eax
131 movl LWP_VMSPACE
(%ecx
), %ecx
/* ECX = vmspace */
132 MPLOCKED btrl
%eax
, VM_PMAP+PM_ACTIVE
(%ecx
)
135 * Push the LWKT switch restore function, which resumes a heavy
136 * weight process. Note that the LWKT switcher is based on
137 * TD_SP, while the heavy weight process switcher is based on
138 * PCB_ESP. TD_SP is usually two ints pushed relative to
139 * PCB_ESP. We push the flags for later restore by cpu_heavy_restore.
142 pushl $cpu_heavy_restore
143 movl
%esp
,TD_SP
(%ebx
)
146 * Save debug regs if necessary
148 movb PCB_FLAGS
(%edx
),%al
150 jz
1f
/* no, skip over */
151 movl
%dr7
,%eax
/* yes, do the save */
152 movl
%eax
,PCB_DR7
(%edx
)
153 andl $
0x0000fc00, %eax
/* disable all watchpoints */
156 movl
%eax
,PCB_DR6
(%edx
)
158 movl
%eax
,PCB_DR3
(%edx
)
160 movl
%eax
,PCB_DR2
(%edx
)
162 movl
%eax
,PCB_DR1
(%edx
)
164 movl
%eax
,PCB_DR0
(%edx
)
169 * Save the FP state if we have used the FP. Note that calling
170 * npxsave will NULL out PCPU(npxthread).
172 cmpl %ebx
,PCPU
(npxthread
)
174 pushl TD_SAVEFPU
(%ebx
)
175 call npxsave
/* do it in a big C function */
176 addl $
4,%esp
/* EAX, ECX, EDX trashed */
178 #endif /* NNPX > 0 */
181 * Switch to the next thread, which was passed as an argument
182 * to cpu_heavy_switch(). Due to the eflags and switch-restore
183 * function we pushed, the argument is at 12(%esp). Set the current
184 * thread, load the stack pointer, and 'ret' into the switch-restore
187 * The switch restore function expects the new thread to be in %eax
188 * and the old one to be in %ebx.
190 * There is a one-instruction window where curthread is the new
191 * thread but %esp still points to the old thread's stack, but
192 * we are protected by a critical section so it is ok.
194 movl
12(%esp
),%eax
/* EAX = newtd, EBX = oldtd */
195 movl
%eax
,PCPU
(curthread
)
196 movl TD_SP
(%eax
),%esp
202 * The switch function is changed to this when a thread is going away
203 * for good. We have to ensure that the MMU state is not cached, and
204 * we don't bother saving the existing thread state before switching.
206 * At this point we are in a critical section and this cpu owns the
207 * thread's token, which serves as an interlock until the switchout is
210 ENTRY
(cpu_exit_switch
)
212 * Get us out of the vmspace
222 movl PCPU
(curthread
),%ebx
225 * If this is a process/lwp, deactivate the pmap after we've
228 movl TD_LWP
(%ebx
),%ecx
231 movl PCPU
(cpuid
), %eax
232 movl LWP_VMSPACE
(%ecx
), %ecx
/* ECX = vmspace */
233 MPLOCKED btrl
%eax
, VM_PMAP+PM_ACTIVE
(%ecx
)
236 * Switch to the next thread. RET into the restore function, which
237 * expects the new thread in EAX and the old in EBX.
239 * There is a one-instruction window where curthread is the new
240 * thread but %esp still points to the old thread's stack, but
241 * we are protected by a critical section so it is ok.
244 movl
%eax
,PCPU
(curthread
)
245 movl TD_SP
(%eax
),%esp
249 * cpu_heavy_restore() (current thread in %eax on entry)
251 * Restore the thread after an LWKT switch. This entry is normally
252 * called via the LWKT switch restore function, which was pulled
253 * off the thread stack and jumped to.
255 * This entry is only called if the thread was previously saved
256 * using cpu_heavy_switch() (the heavy weight process thread switcher),
257 * or when a new process is initially scheduled. The first thing we
258 * do is clear the TDF_RUNNING bit in the old thread and set it in the
261 * NOTE: The lwp may be in any state, not necessarily LSRUN, because
262 * a preemption switch may interrupt the process and then return via
265 * YYY theoretically we do not have to restore everything here, a lot
266 * of this junk can wait until we return to usermode. But for now
267 * we restore everything.
269 * YYY the PCB crap is really crap, it makes startup a bitch because
270 * we can't switch away.
272 * YYY note: spl check is done in mi_switch when it splx()'s.
275 ENTRY
(cpu_heavy_restore
)
277 movl TD_PCB
(%eax
),%edx
/* EDX = PCB */
278 movl TD_LWP
(%eax
),%ecx
280 #if defined(SWTCH_OPTIM_STATS)
281 incl _swtch_optim_stats
284 * Tell the pmap that our cpu is using the VMSPACE now. We cannot
285 * safely test/reload %cr3 until after we have set the bit in the
286 * pmap (remember, we do not hold the MP lock in the switch code).
288 movl LWP_VMSPACE
(%ecx
), %ecx
/* ECX = vmspace */
289 movl PCPU
(cpuid
), %esi
290 MPLOCKED btsl
%esi
, VM_PMAP+PM_ACTIVE
(%ecx
)
293 * Restore the MMU address space. If it is the same as the last
294 * thread we don't have to invalidate the tlb (i.e. reload cr3).
295 * YYY which naturally also means that the PM_ACTIVE bit had better
296 * already have been set before we set it above, check? YYY
300 movl PCB_CR3
(%edx
),%ecx
303 #if defined(SWTCH_OPTIM_STATS)
304 decl _swtch_optim_stats
305 incl _tlb_flush_count
311 * Clear TDF_RUNNING flag in old thread only after cleaning up
312 * %cr3. The target thread is already protected by being TDF_RUNQ
313 * so setting TDF_RUNNING isn't as big a deal.
315 andl $~TDF_RUNNING
,TD_FLAGS
(%ebx
)
316 orl $TDF_RUNNING
,TD_FLAGS
(%eax
)
320 * Deal with the PCB extension, restore the private tss
322 movl PCB_EXT
(%edx
),%edi
/* check for a PCB extension */
323 movl $
1,%ebx
/* maybe mark use of a private tss */
328 * Going back to the common_tss. We may need to update TSS_ESP0
329 * which sets the top of the supervisor stack when entering from
330 * usermode. The PCB is at the top of the stack but we need another
331 * 16 bytes to take vm86 into account.
334 movl
%ebx
, PCPU
(common_tss
) + TSS_ESP0
336 cmpl $
0,PCPU
(private_tss
) /* don't have to reload if */
337 je
3f
/* already using the common TSS */
339 subl
%ebx
,%ebx
/* unmark use of private tss */
342 * Get the address of the common TSS descriptor for the ltr.
343 * There is no way to get the address of a segment-accessed variable
344 * so we store a self-referential pointer at the base of the per-cpu
345 * data area and add the appropriate offset.
347 movl $gd_common_tssd
, %edi
351 * Move the correct TSS descriptor into the GDT slot, then reload
355 movl
%ebx
,PCPU
(private_tss
) /* mark/unmark private tss */
356 movl PCPU
(tss_gdt
), %ebx
/* entry in GDT */
361 movl $GPROC0_SEL
*8, %esi
/* GSEL(entry, SEL_KPL) */
366 * Restore general registers.
368 movl PCB_EBX
(%edx
),%ebx
369 movl PCB_ESP
(%edx
),%esp
370 movl PCB_EBP
(%edx
),%ebp
371 movl PCB_ESI
(%edx
),%esi
372 movl PCB_EDI
(%edx
),%edi
373 movl PCB_EIP
(%edx
),%eax
378 * Restore the user LDT if we have one
380 cmpl $
0, PCB_USERLDT
(%edx
)
382 movl _default_ldt
,%eax
383 cmpl PCPU
(currentldt
),%eax
386 movl
%eax
,PCPU
(currentldt
)
395 * Restore the user TLS if we have one
403 * Restore the DEBUG register state if necessary.
405 movb PCB_FLAGS
(%edx
),%al
407 jz
1f
/* no, skip over */
408 movl PCB_DR6
(%edx
),%eax
/* yes, do the restore */
410 movl PCB_DR3
(%edx
),%eax
412 movl PCB_DR2
(%edx
),%eax
414 movl PCB_DR1
(%edx
),%eax
416 movl PCB_DR0
(%edx
),%eax
418 movl
%dr7
,%eax
/* load dr7 so as not to disturb */
419 andl $
0x0000fc00,%eax
/* reserved bits */
421 movl PCB_DR7
(%edx
),%ebx
422 andl $~
0x0000fc00,%ebx
433 * Update pcb, saving current processor state.
439 /* caller's return address - child won't execute this routine */
441 movl
%eax
,PCB_EIP
(%ecx
)
442 movl
%ebx
,PCB_EBX
(%ecx
)
443 movl
%esp
,PCB_ESP
(%ecx
)
444 movl
%ebp
,PCB_EBP
(%ecx
)
445 movl
%esi
,PCB_ESI
(%ecx
)
446 movl
%edi
,PCB_EDI
(%ecx
)
450 * If npxthread == NULL, then the npx h/w state is irrelevant and the
451 * state had better already be in the pcb. This is true for forks
452 * but not for dumps (the old book-keeping with FP flags in the pcb
453 * always lost for dumps because the dump pcb has 0 flags).
455 * If npxthread != NULL, then we have to save the npx h/w state to
456 * npxthread's pcb and copy it to the requested pcb, or save to the
457 * requested pcb and reload. Copying is easier because we would
458 * have to handle h/w bugs for reloading. We used to lose the
459 * parent's npx state for forks by forgetting to reload.
461 movl PCPU
(npxthread
),%eax
465 pushl
%ecx
/* target pcb */
466 movl TD_SAVEFPU
(%eax
),%eax
/* originating savefpu area */
476 pushl $PCB_SAVEFPU_SIZE
477 leal PCB_SAVEFPU
(%ecx
),%ecx
482 #endif /* NNPX > 0 */
488 * cpu_idle_restore() (current thread in %eax on entry) (one-time execution)
490 * Don't bother setting up any regs other then %ebp so backtraces
491 * don't die. This restore function is used to bootstrap into the
492 * cpu_idle() LWKT only, after that cpu_lwkt_*() will be used for
495 * Clear TDF_RUNNING in old thread only after we've cleaned up %cr3.
497 * If we are an AP we have to call ap_init() before jumping to
498 * cpu_idle(). ap_init() will synchronize with the BP and finish
499 * setting up various ncpu-dependant globaldata fields. This may
500 * happen on UP as well as SMP if we happen to be simulating multiple
503 ENTRY
(cpu_idle_restore
)
507 andl $~TDF_RUNNING
,TD_FLAGS
(%ebx
)
508 orl $TDF_RUNNING
,TD_FLAGS
(%eax
)
519 * cpu_kthread_restore() (current thread is %eax on entry) (one-time execution)
521 * Don't bother setting up any regs other then %ebp so backtraces
522 * don't die. This restore function is used to bootstrap into an
523 * LWKT based kernel thread only. cpu_lwkt_switch() will be used
526 * Since all of our context is on the stack we are reentrant and
527 * we can release our critical section and enable interrupts early.
529 ENTRY
(cpu_kthread_restore
)
531 movl TD_PCB
(%eax
),%edx
533 andl $~TDF_RUNNING
,TD_FLAGS
(%ebx
)
534 orl $TDF_RUNNING
,TD_FLAGS
(%eax
)
535 subl $TDPRI_CRIT
,TD_PRI
(%eax
)
536 popl
%eax
/* kthread exit function */
537 pushl PCB_EBX
(%edx
) /* argument to ESI function */
538 pushl
%eax
/* set exit func as return address */
539 movl PCB_ESI
(%edx
),%eax
545 * Standard LWKT switching function. Only non-scratch registers are
546 * saved and we don't bother with the MMU state or anything else.
548 * This function is always called while in a critical section.
550 * There is a one-instruction window where curthread is the new
551 * thread but %esp still points to the old thread's stack, but
552 * we are protected by a critical section so it is ok.
556 ENTRY
(cpu_lwkt_switch
)
557 pushl
%ebp
/* note: GDB hacked to locate ebp relative to td_sp */
559 movl PCPU
(curthread
),%ebx
563 /* warning: adjust movl into %eax below if you change the pushes */
567 * Save the FP state if we have used the FP. Note that calling
568 * npxsave will NULL out PCPU(npxthread).
570 * We have to deal with the FP state for LWKT threads in case they
571 * happen to get preempted or block while doing an optimized
572 * bzero/bcopy/memcpy.
574 cmpl %ebx
,PCPU
(npxthread
)
576 pushl TD_SAVEFPU
(%ebx
)
577 call npxsave
/* do it in a big C function */
578 addl $
4,%esp
/* EAX, ECX, EDX trashed */
580 #endif /* NNPX > 0 */
582 movl
4+20(%esp
),%eax
/* switch to this thread */
583 pushl $cpu_lwkt_restore
584 movl
%esp
,TD_SP
(%ebx
)
585 movl
%eax
,PCPU
(curthread
)
586 movl TD_SP
(%eax
),%esp
589 * eax contains new thread, ebx contains old thread.
594 * cpu_lwkt_restore() (current thread in %eax on entry)
596 * Standard LWKT restore function. This function is always called
597 * while in a critical section.
599 * Warning: due to preemption the restore function can be used to
600 * 'return' to the original thread. Interrupt disablement must be
601 * protected through the switch so we cannot run splz here.
603 ENTRY
(cpu_lwkt_restore
)
604 andl $~TDF_RUNNING
,TD_FLAGS
(%ebx
)
605 orl $TDF_RUNNING
,TD_FLAGS
(%eax
)
616 * Make AP become the idle loop.
618 ENTRY
(bootstrap_idle
)
619 movl PCPU
(curthread
),%eax
621 movl TD_SP
(%eax
),%esp